NAT10 Is Involved in Cardiac Remodeling Through ac4C-Mediated Transcriptomic Regulation.

BACKGROUND: Heart failure, characterized by cardiac remodeling, is associated with abnormal epigenetic processes and aberrant gene expression. Here, we aimed to elucidate the effects and mechanisms of NAT10 (N-acetyltransferase 10)-mediated N4-acetylcytidine (ac4C) acetylation during cardiac remodeling. METHODS: NAT10 and ac4C expression were detected in both human and mouse subjects with cardiac remodeling through multiple assays. Subsequently, acetylated RNA immunoprecipitation and sequencing, thiol-linked alkylation for the metabolic sequencing of RNA (SLAM-seq), and ribosome sequencing (Ribo-seq) were employed to elucidate the role of ac4C-modified posttranscriptional regulation in cardiac remodeling. Additionally, functional experiments involving the overexpression or knockdown of NAT10 were conducted in mice models challenged with Ang II (angiotensin II) and transverse aortic constriction. RESULTS: NAT10 expression and RNA ac4C levels were increased in in vitro and in vivo cardiac remodeling models, as well as in patients with cardiac hypertrophy. Silencing and inhibiting NAT10 attenuated Ang II-induced cardiomyocyte hypertrophy and cardiofibroblast activation. Next-generation sequencing revealed ac4C changes in both mice and humans with cardiac hypertrophy were associated with changes in global mRNA abundance, stability, and translation efficiency. Mechanistically, NAT10 could enhance the stability and translation efficiency of CD47 and ROCK2 transcripts by upregulating their mRNA ac4C modification, thereby resulting in an increase in their protein expression during cardiac remodeling. Furthermore, the administration of Remodelin, a NAT10 inhibitor, has been shown to prevent cardiac functional impairments in mice subjected to transverse aortic constriction by suppressing cardiac fibrosis, hypertrophy, and inflammatory responses, while also regulating the expression levels of CD47 and ROCK2 (Rho associated coiled-coil containing protein kinase 2). CONCLUSIONS: Therefore, our data suggest that modulating epitranscriptomic processes, such as ac4C acetylation through NAT10, may be a promising therapeutic target against cardiac remodeling.

Spectroscopic measurement of the two-dimensional flame temperature based on a perovskite single photodetector.

Existing non-contact flame temperature measuring methods depend on complex, bulky and expensive optical instruments, which make it difficult for portable applications and high-density distributed networking monitoring. Here, we demonstrate a flame temperature imaging method based on a perovskite single photodetector. High-quality perovskite film epitaxy grows on the SiO2/Si substrate to fabricate the photodetector. Duo to the Si/MAPbBr3 heterojunction, the light detection wavelength is extended from 400 nm to 900 nm. Then, a perovskite single photodetector spectrometer has been developed using the deep-learning method for spectroscopic measurement of flame temperature. In the temperature test experiment, the spectral line of doping element K+ has been selected to measure the flame temperature. The photoresponsivity function of the wavelength was learned based on a commercial standard blackbody source. The spectral line of element K+ has been reconstructed using the photocurrents matrix by the regression solving photoresponsivity function. As a validation experiment, the "NUC" pattern is realized by scanning the perovskite single-pixel photodetector. Finally, the flame temperature of adulterated element K+ has been imaged with the error of 5%. It provides a way to develop high precision, portable, low-cost flame temperature imaging technology.

Perovskite single-detector visible-light spectrometer.

We demonstrate a perovskite single-phototransistor visible-light spectrometer based on a deep-learning method. The size of the spectrometer is set to the scale of the phototransistor. A photoresponsivity matrix for the deep-learning system is learned from the characteristic parameters of the visible-light wavelength, gate voltage, and power densities of a commercial standard blackbody source. Unknown spectra are reconstructed using the corresponding photocurrent vectors. As a confirmatory experiment, a 532-nm laser and multipeak broadband spectrum are successfully reconstructed using our perovskite single-phototransistor spectrometer. The resolution is improved to 1 nm by increasing the number of sampling points from 80 to 400. In addition, a way to further improve the resolution is provided by increasing the number of sampling points, characteristic parameters, and training datasets. Furthermore, artificial intelligence technology may open pathways for on-chip visible-light spectroscopy.

Frequency of prothrombin time-international normalized ratio monitoring and the long-term prognosis in patients with mechanical valve replacement.

BACKGROUND: The study aimed to assess the correlation between the monitoring frequency of PT-INR and the long-term prognosis in patients with mechanical heart valve (MHV) replacement after discharge. METHODS: This single-center, observational study enrolled patients who underwent MHV replacement and discharged from June 2015 to May 2018. Patients or their corresponding family members were followed with a telephone questionnaire survey in July-October 2020. Based on monitoring intervals, patients were divided into frequent monitoring (FM) group (≤ 1 month) and less frequent monitoring (LFM) group (> 1 month). The primary endpoint was the composite of thromboembolic event, major bleeding or all-cause death. The secondary endpoints were thromboembolic event, major bleeding or all-cause death, respectively. RESULTS: A total of 188 patients were included in the final analysis. The median follow-up duration was 3.6 years (Interquartile range: 2.6 to 4.4 years). 104 (55.3%) patients and 84 (44.7%) patients were classified into the FM group and the LFM group, respectively. The FM group had a significantly lower incidence of the primary endpoint than the LFM group (3.74 vs. 1.16 per 100 patient-years, adjusted HR: 3.31 [95% CI 1.05-10.42, P = 0.041]). Secondary analysis revealed that the risk of thromboembolic events and all-cause death were also reduced in the FM group. CONCLUSIONS: The management of warfarin treatment in patients after MHV replacement remains challenging. Patients with less frequent monitoring of PT-INR might have worse clinical prognosis than those with frequent PT-INR monitoring.

Transcription factor Sp2 promotes TGFB-mediated interstitial cell osteogenic differentiation in bicuspid aortic valves through a SMAD-dependent pathway.

Calcification of the bicuspid aortic valve (BAV) involves differential expression of various RNA genes, which is achieved through complex regulatory networks that are controlled in part by transcription factors and microRNAs. We previously found that miR-195-5p regulates the osteogenic differentiation of valvular interstitial cells (VICs) by targeting the TGF-ß pathway. However, the transcriptional regulation of miR-195-5p in calcified BAV patients is not yet clear. In this study, stenotic aortic valve tissues from patients with BAVs and tricuspid aortic valves (TAVs) were collected. Candidate transcription factors of miR-195-5p were predicted by bioinformatics analysis and tested in diseased valves and in male porcine VICs. SP2 gene expression and the corresponding protein levels in BAV were significantly lower than those in TAV, and a low SP2 expression level environment in VICs resulted in remarkable increases in RNA expression levels of RUNX2, BMP2, collagen 1, MMP2, and MMP9 and the corresponding proteins. ChIP assays revealed that SP2 directly bound to the transcription promoter region of miR-195-5p. Cotransfection of SP2 shRNA and a miR-195-5p mimic in porcine VICs demonstrated that SP2 repressed SMAD7 expression via miR-195-5p, while knockdown of SP2 increased the mRNA expression of SMAD7 and the corresponding protein and attenuated Smad 2/3 expression. Immunofluorescence staining of diseased valves confirmed that the functional proteins of osteogenesis differentiation, including RUNX2, BMP2, collagen 1, and osteocalcin, were overexpressed in BAVs. In Conclusion, the transcription factor Sp2 is expressed at low levels in VICs from BAV patients, which has a negative impact on miR-195-5p expression by binding its promoter region and partially promotes calcification through a SMAD-dependent pathway.

Surgical ablation supplemented by ethanol injection for ventricular tachycardia refractory to percutaneous ablation.

BACKGROUND: A combination of endocardial and epicardial approaches has improved the overall success rate of ventricular tachycardia (VT) ablation in patients with cardiomyopathy. However, the origins of some VTs are truly intramural or close to coronary arteries, which makes this combined strategy either prone to failure or too risky. OBJECTIVES: This observational study aimed to explore the feasibility and efficacy of direct epicardial ablation combined with intramural ethanol injection via surgical approach for inaccessible intramural VTs or VTs too close to coronary arteries. METHODS: In four canines ventricular lesions produced by direct epicardial injection of ethanol were assessed. Six consecutive patients with recurrent VT refractory to catheter endocardial and epicardial RF ablation and that remained inducible after surgical epicardial mapping and RF ablation were included. Ethanol was injected by needle at the epicardial RF ablation sites. The primary outcome was freedom of sustained VT determined by device interrogation and periodical 24-h holter recordings subsequently. RESULTS: In an animal study, the lesions were homogenous and increased in size with the volume of ethanol injected. In all six patients, ethanol injection at the target sites in the anterior or lateral left ventricle abolished inducible VT. Over a median follow-up of 22 months (range, 6-65), all patients remained free of sustained VT. One patient died of pulmonary infection one year after the procedure. CONCLUSIONS: A hybrid strategy of surgical ablation combined with intramural ethanol injection is feasible and effective in patients with multiple failed percutaneous ablation attempts.

Structural Transformation of 2,7-Dibromopyrene on Au(111) Mediated by Halogen-Bonding Motifs.

Fundamental understanding of the bonding motifs that elaborately mediate the formation of supramolecular nanostructures is essential for the rational design of stable artificial organic architectures. Herein, the structural transformation of the adsorption complex of 2, 7-dibromopyrene (Br2 Py) on the Au(111) surface has been investigated by scanning tunnelling microscopy combined with X-ray photoemission spectroscopy and density function theory calculations. In the initial stage of self-assembly, well ordered patterns are formed in the manner of extended supramolecular structures balanced by intermolecular halogen bonding motifs, whilst the Au(111) reconstruction is still fairly visible. Subsequent thermal annealing promotes the dehalogenation and on-surface Ullmann coupling, and polymerized oligomers are consequently constructed. Interestingly, such polymerized chains are still stably mediated by the halogen bonding motif via dissociated Br atoms which are revealed to be absorbed on the bridge site of Au(111), while the number of halogen bonds increases significantly from self-assembly to Ullmann coupling polymerization, indicating that the halogen bonding motif contributes significantly to the extended one-dimensional polymers.

Integrated Bioinformatics Analysis Predicts the Key Genes Involved in Aortic Valve Calcification: From Hemodynamic Changes to Extracellular Remodeling.

In our aging world, increasing numbers of people are suffering from calcific aortic valve disease (CAVD). In this study, we used integrated bioinformatics analysis to predict several key genes that are involved in the initiation and progression of CAVD. Expression profiles of 15 calcific and 14 normal human aortic valve samples were generated from two gene expression datasets (GSE12644 and GSE51472). Dataset GSE26953 from the human aortic valve fibrosa-derived endothelial cells cultured under laminar or oscillatory shear stress was also evaluated. Related R packages were used to process the data. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed for functional annotation. Hub genes were identified based on the protein-protein interaction network. CAVD-related gene modules were identified by Weighted Gene Co-expression Network Analysis (WGCNA). The predicted key genes were manually reviewed. In our present work, complex connections among mechano-response, oxidative stress, inflammation and extracellular remodeling pathways in the etiology of CAVD were revealed. The key genes, thus identified, encode a transcription factor KLF2 and phospholipid phosphatase 3 (PLPP3) that are involved in mechano-responses; eNOS involved in oxidative stress; IL-8 involved in inflammation; and collagen triple helix repeat containing 1 (CTHRC1) and secretogranin II (SCG2) involved in extracellular remodeling. These gene products are predicted to play critical roles in CAVD development and progression. The present study provides valuable information for future research and drug development.

MicroRNA-19a-3p augments TGF-ß1-induced cardiac fibroblast activation via targeting BAMBI.

The main pathogenic factor leading to cardiac remodeling and heart failure is myocardial fibrosis. Recent research indicates that microRNAs are essential for the progress of cardiac fibrosis. Myocardial fibrosis is considered to be alleviated through the bone morphogenetic protein and activin membrane-bound inhibitor (BAMBI), which does this by blocking the transforming growth factor ß1 (TGF-ß1) signaling pathway. Here, this study sought to elucidate the post-transcriptional regulation of miR-19a-3p on BAMBI and its role in TGF-ß1-induced cardiac fibroblast activation. Transverse aortic constriction (TAC) caused both myocardial interstitial and perivascular collagen deposition. RT-PCR showed that miR-19a-3p was upregulated in the myocardial tissue of cardiac fibrosis, and TGF-ß1 induced an increase of miR-19a-3p expression in cardiac fibroblasts. The dual-luciferase reporter test and qRT-PCR confirmed that miR-19a-3p directly combined with BAMBI mRNA 3'UTR, thus reduced BAMBI expression, which diminished the capability of BAMBI to inhibit TGF-ß1. Furthermore, miR-19a-3p mimic increased the activation of TGF-ß1/SMAD2/3 pathway signaling, which supported cardiac fibroblast activation, which blocked by overexpression of BAMBI. These findings imply that miR-19a-3p enhances the activation of TGF-ß1/SMAD2/3 by inhibiting BAMBI, further boosting the activation of cardiac fibroblasts, and may thus offer a novel strategy to tackling myocardial fibrosis.

Optoelectronic synapses based on a triple cation perovskite and Al/MoO3 interface for neuromorphic information processing.

Optoelectronic synaptic transistors are attractive for applications in next-generation brain-like computation systems, especially for their visible-light operation and in-sensor computing capabilities. However, from a material perspective, it is difficult to build a device that meets expectations in terms of both its functions and power consumption, prompting the call for greater innovation in materials and device construction. In this study, we innovatively combined a novel perovskite carrier supply layer with an Al/MoO3 interface carrier regulatory layer to fabricate optoelectronic synaptic devices, namely Al/MoO3/CsFAMA/ITO transistors. The device could mimic a variety of biological synaptic functions and required ultralow-power consumption during operation with an ultrafast speed of >0.1 µs under an optical stimulus of about 3 fJ, which is equivalent to biological synapses. Moreover, Pavlovian conditioning and visual perception tasks could be implemented using the spike-number-dependent plasticity (SNDP) and spike-rate-dependent plasticity (SRDP). This study suggests that the proposed CsFAMA synapse with an Al/MoO3 interface has the potential for ultralow-power neuromorphic information processing.

Checkpoint Kinase 1 Stimulates Endogenous Cardiomyocyte Renewal and Cardiac Repair by Binding to Pyruvate Kinase Isoform M2 C-Domain and Activating Cardiac Metabolic Reprogramming in a Porcine Model of Myocardial Ischemia/Reperfusion Injury.

BACKGROUND: The regenerative capacity of the adult mammalian hearts is limited. Numerous studies have explored mechanisms of adult cardiomyocyte cell-cycle withdrawal. This translational study evaluated the effects and underlying mechanism of rhCHK1 (recombinant human checkpoint kinase 1) on the survival and proliferation of cardiomyocyte and myocardial repair after ischemia/reperfusion injury in swine. METHODS AND RESULTS: Intramyocardial injection of rhCHK1 protein (1 mg/kg) encapsulated in hydrogel stimulated cardiomyocyte proliferation and reduced cardiac inflammation response at 3 days after ischemia/reperfusion injury, improved cardiac function and attenuated ventricular remodeling, and reduced the infarct area at 28 days after ischemia/reperfusion injury. Mechanistically, multiomics sequencing analysis demonstrated enrichment of glycolysis and mTOR (mammalian target of rapamycin) pathways after rhCHK1 treatment. Co-Immunoprecipitation (Co-IP) experiments and protein docking prediction showed that CHK1 (checkpoint kinase 1) directly bound to and activated the Serine 37 (S37) and Tyrosine 105 (Y105) sites of PKM2 (pyruvate kinase isoform M2) to promote metabolic reprogramming. We further constructed plasmids that knocked out different CHK1 and PKM2 amino acid domains and transfected them into Human Embryonic Kidney 293T (HEK293T) cells for CO-IP experiments. Results showed that the 1-265 domain of CHK1 directly binds to the 157-400 amino acids of PKM2. Furthermore, hiPSC-CM (human iPS cell-derived cardiomyocyte) in vitro and in vivo experiments both demonstrated that CHK1 stimulated cardiomyocytes renewal and cardiac repair by activating PKM2 C-domain-mediated cardiac metabolic reprogramming. CONCLUSIONS: This study demonstrates that the 1-265 amino acid domain of CHK1 binds to the 157-400 domain of PKM2 and activates PKM2-mediated metabolic reprogramming to promote cardiomyocyte proliferation and myocardial repair after ischemia/reperfusion injury in adult pigs.

Effect of extracorporeal shock wave combined with pregabalin on patients with post-herpetic neuralgia.

Post-herpetic neuralgia (PHN) is a chronic pain that is difficult to treat and lasts a long time, which poses a threat to patients' physical and mental health (MH) and quality of life. To analyze the effectiveness of extracorporeal shock wave (ESW) combined with pregabalin on PHN and its impact on PHN patients' quality of life with the help of a random number table. Totally 164 PHN patients were assigned to a control group (n = 82) or an observation group (n = 82). The observation group was given pregabalin combined with ESW treatment, while the control group was only given pregabalin. In the 2 groups, the general clinical data of the patients were compared. The inflammation levels including erythrocyte sedimentation rate (ESR), CRP, lymphocyte count and albumin level in both groups were compared prior to and following therapy. In addition, the difference between pretreatment and post-treatment in the 2 groups was compared with respect to neuralgia and quality of life. After treatment, the observation group exhibited much lower ESR and CRP but quite higher lymphocyte count and albumin level relative to the control group (P < .05). Additionally, the improvement in visual analogue scale (VAS) scores and short form 36 (SF-36) scale scores including role-emotional (RE), MH, physical function (PF), general health (GH), bodily pain (BP), social function (SF), vitality (VT), and role-physical (RP) scores in the 2 groups were pronounced following treatment, which was more apparent in the observation group (P < .05). The combination of ESW with pregabalin can reduce the inflammation, improve the quality of life of PHN patients and effectively relieve their neuralgia.

Towards Accurate and Robust Domain Adaptation Under Multiple Noisy Environments.

In many non-stationary environments, machine learning algorithms usually confront the distribution shift scenarios. Previous domain adaptation methods have achieved great success. However, they would lose algorithm robustness in multiple noisy environments where the examples of source domain become corrupted by label noise, feature noise, or open-set noise. In this paper, we report our attempt toward achieving noise-robust domain adaptation. We first give a theoretical analysis and find that different noises have disparate impacts on the expected target risk. To eliminate the effect of source noises, we propose offline curriculum learning minimizing a newly-defined empirical source risk. We suggest a proxy distribution-based margin discrepancy to gradually decrease the noisy distribution distance to reduce the impact of source noises. We propose an energy estimator for assessing the outlier degree of open-set-noise examples to defeat the harmful influence. We also suggest robust parameter learning to mitigate the negative effect further and learn domain-invariant feature representations. Finally, we seamlessly transform these components into an adversarial network that performs efficient joint optimization for them. A series of empirical studies on the benchmark datasets and the COVID-19 screening task show that our algorithm remarkably outperforms the state-of-the-art, with over 10% accuracy improvements in some transfer tasks.

Biomechanical characterization of normal and pathological human ascending aortic tissues via biaxial testing Experiment, constitutive modeling and finite element analysis.

BACKGROUND: Aortic dissection and atherosclerosis are two common pathological conditions affecting the aorta. Aortic biomechanics are believed to be closely associated with the pathological development of these diseases. However, the biomechanical environment that predisposes the aortic wall to these pathological conditions remains unclear. METHODS: Sixteen ascending aortic specimens were harvested from 16 human subjects and further categorized into three groups according to their disease states: aortic dissection group, aortic dissection with accompanied atherosclerosis group and healthy group. Experimental stress-strain data from biaxial tensile testing were used to fit the anisotropic Mooney-Rivlin model to determine material parameters. Computed tomography images or transesophageal echocardiography images were collected to construct computational models to simulate the stress/strain distributions in aortas at the pre-dissection state. Statistical analyses were performed to identify the biomechanical factors to distinguish three groups of aortic tissues. RESULTS: Material parameters of anisotropic Mooney-Rivlin model were fitted with average R2 value 0.9749. The aortic diameter showed no significant difference among three groups. Changes of maximum and average stress values from minimum pressure to maximum pressure (△MaxStress and △AveStress) had significantly difference between dissection group and dissection with accompanied atherosclerosis group (p = 0.0201 and 0.0102). Changes of maximum and average strain values from minimum pressure to maximum pressure (△MaxStrain and △AveStrain) from dissection group were significant different from healthy group (p = 0.0171 and 0.0281). CONCLUSION: Changes of stress and strain values during the cardiac cycle are good biomechanical factors for predicting potential aortic dissection and aortic dissection accompanied with atherosclerosis.

Improved methodology for efficient establishment of the myocardial ischemia-reperfusion model in pigs through the median thoracic incision.

To investigate the feasibility and effectiveness of establishing porcine ischemia-reperfusion models by ligating the left anterior descending (LAD) coronary artery, we first randomly divided 16 male Bama pigs into a sham group and a model group. After anesthesia, we separated the arteries and veins. Subsequently, we rapidly located the LAD coronary artery at the beginning of its first diagonal branch through a mid-chest incision. Then, we loosened and released the ligation line after five minutes of pre-occlusion. Finally, we ligated the LAD coronary artery in situ two minutes later and loosened the ligature 60 min after ischemia. Compared with the sham group, electrocardiogram showed multiple continuous lead ST-segment elevations, and ultrasound cardiogram showed significantly lower ejection fraction and left ventricular fractional shortening at one hour and seven days post-operation in the model group. Twenty-four hours after the operation, cardiac troponin T and creatine kinase-MB isoenzyme levels significantly increased in the model group, compared with the sham group. Hematoxylin and eosin staining showed the presence of many inflammatory cells infiltrating the interstitium of the myocardium in the model group but not in the sham group. Masson staining revealed a significant increase in infarct size in the ischemia/reperfusion group. All eight pigs in the model group recovered with normal sinus heart rates, and the survival rate was 100%. In conclusion, the method can provide an accurate and stable large animal model for preclinical research on ischemia/reperfusion with a high success rate and homogeneity of the myocardial infarction area.

Sleeve lobectomy versus pneumonectomy for non-small cell lung cancer: a meta-analysis.

AIM: It is controversial that whether sleeve lobectomy (SL) should be promoted more worthy than pneumonectomy (PN) in suitable patients. METHODS: We searched all studies that had been published in English from PUBMED and Embase which compared the short-term and long-term outcomes of SL and pneumonectomy (PN) in patients with non-small cell lung cancer (NSCLC). RESULTS: Nineteen studies met our criteria with a combined total of 3878 subjects, of which 1316 (33.9%) underwent SL and 2562 (66.1%) underwent PN. The odds ratio was 0.50 (95% CI: 0.34-0.72) for postoperative mortality, 1.17 (95% CI: 0.82-1.67) for postoperative complications, 0.78 (95% CI: 0.47-1.29) for locoregional recurrences. The risk difference for 1-, 3-, 5- year was 0.11 (95% CI: 0.07-0.14), 0.15 (95% CI: 0.06-0.24), 0.15 (95% CI: 0.09-0.20),respectively. The pooled hazard ratio was 0.63 (95% CI: 0.56-0.71) in favor of SL group. CONCLUSION: SL is more worthy to be done than PN in suitable patients with less mortality and better long-term survival.

Effects of Ti Target Purity and Microstructure on Deposition Rate, Microstructure and Properties of Ti Films.

Three titanium (Ti) targets with different purities were used to prepare Ti films on polyimide substrates by DC magnetron sputtering. The microstructures of Ti films were characterized by a metallographic microscope, X-ray diffractometer, field emission scanning electron microscope and three-dimensional surface topography instrument. In this study, we investigated the effects of Ti target purity and microstructure on film deposition rate, surface roughness, microstructure and resistivity. The results show that the deposition rate increased with increasing Ti target purity. Ti film deposited by the high-purity (99.999%) Ti target has fewer surface particles with smaller size, lower surface roughness and lower resistivity when compared to that prepared by the Ti target of low purity (99.7%). The surface roughness of Ti film prepared by the high-purity Ti target was Sa = 121 nm, the deposition rate was 16.3 nm/min and the resistivity was 6.9 × 10-6 Ω·m. For Ti targets of the same purity, the performance of Ti film prepared by a target with equiaxed α-phase grains is better than that of Ti film prepared by a target with twins and ß-phase grains.

Learning Transferable Parameters for Unsupervised Domain Adaptation.

Unsupervised domain adaptation (UDA) enables a learning machine to adapt from a labeled source domain to an unlabeled target domain under the distribution shift. Thanks to the strong representation ability of deep neural networks, recent remarkable achievements in UDA resort to learning domain-invariant features. Intuitively, the goal is that a good feature representation and the hypothesis learned from the source domain can generalize well to the target domain. However, the learning processes of domain-invariant features and source hypotheses inevitably involve domain-specific information that would degrade the generalizability of UDA models on the target domain. The lottery ticket hypothesis proves that only partial parameters are essential for generalization. Motivated by it, we find in this paper that only partial parameters are essential for learning domain-invariant information. Such parameters are termed transferable parameters that can generalize well in UDA. In contrast, the rest parameters tend to fit domain-specific details and often cause the failure of generalization, which are termed untransferable parameters. Driven by this insight, we propose Transferable Parameter Learning (TransPar) to reduce the side effect of domain-specific information in the learning process and thus enhance the memorization of domain-invariant information. Specifically, according to the distribution discrepancy degree, we divide all parameters into transferable and untransferable ones in each training iteration. We then perform separate update rules for the two types of parameters. Extensive experiments on image classification and regression tasks (keypoint detection) show that TransPar outperforms prior arts by non-trivial margins. Moreover, experiments demonstrate that TransPar can be integrated into the most popular deep UDA networks and be easily extended to handle any data distribution shift scenarios.

MetaKernel: Learning Variational Random Features with Limited Labels.

Few-shot learning deals with the fundamental and challenging problem of learning from a few annotated samples, while being able to generalize well on new tasks. The crux of few-shot learning is to extract prior knowledge from related tasks to enable fast adaptation to a new task with a limited amount of data. In this paper, we propose meta-learning kernels with random Fourier features for few-shot learning, we call MetaKernel. Specically, we propose learning variational random features in a data-driven manner to obtain task-specic kernels by leveraging the shared knowledge provided by related tasks in a meta-learning setting. We treat the random feature basis as the latent variable, which is estimated by variational inference. The shared knowledge from related tasks is incorporated into a context inference of the posterior, which we achieve via a long-short term memory module. To establish more expressive kernels, we deploy conditional normalizing ows based on coupling layers to achieve a richer posterior distribution over random Fourier bases. The resultant kernels are more informative and discriminative, which further improves the few-shot learning. We conduct experiments on both few-shot image classication and regression tasks. The results on fourteen datasets demonstrate MetaKernel consistently better performance than state-of-the-art alternatives.

Flame Edge Detection Method Based on a Convolutional Neural Network.

In this study, an improved flame edge detector based on convolutional neural network (CNN) was proposed. The proposed method can generate edge graphs and extract edge graphs relatively effectively. Our network architecture was based on VGG16 primarily, the last two max-pooling operators and all full connection layers of the VGG16 network were deleted, and the rest was taken as the basic network. The images output by the five convolution layers were upsampled to the size of the input images and finally fused to the edge image. Error calculation and back propagation of the fusion image and label image are carried out to form a weakly supervised model. Using the open datasets BSDS500 to train the network, the ODS F-measure can reach 0.810. Various experiments were carried out on different flame and fire images, including butane-air flame, oxygen-ethanol flame, energetic material flame, and oxygen-acetylene premixed jet flame, and the infrared thermogram was also verified by our method. The results demonstrate the effectiveness and robustness of the proposed algorithm.